michael@0: /*
michael@0:  * Copyright © 2009  Red Hat, Inc.
michael@0:  * Copyright © 2009  Keith Stribley
michael@0:  * Copyright © 2011  Google, Inc.
michael@0:  *
michael@0:  *  This is part of HarfBuzz, a text shaping library.
michael@0:  *
michael@0:  * Permission is hereby granted, without written agreement and without
michael@0:  * license or royalty fees, to use, copy, modify, and distribute this
michael@0:  * software and its documentation for any purpose, provided that the
michael@0:  * above copyright notice and the following two paragraphs appear in
michael@0:  * all copies of this software.
michael@0:  *
michael@0:  * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
michael@0:  * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
michael@0:  * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
michael@0:  * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
michael@0:  * DAMAGE.
michael@0:  *
michael@0:  * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
michael@0:  * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
michael@0:  * FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
michael@0:  * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
michael@0:  * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
michael@0:  *
michael@0:  * Red Hat Author(s): Behdad Esfahbod
michael@0:  * Google Author(s): Behdad Esfahbod
michael@0:  */
michael@0: 
michael@0: #include "hb-private.hh"
michael@0: 
michael@0: #include "hb-icu.h"
michael@0: 
michael@0: #include "hb-unicode-private.hh"
michael@0: 
michael@0: #include <unicode/uchar.h>
michael@0: #include <unicode/unorm.h>
michael@0: #include <unicode/ustring.h>
michael@0: #include <unicode/uversion.h>
michael@0: 
michael@0: 
michael@0: hb_script_t
michael@0: hb_icu_script_to_script (UScriptCode script)
michael@0: {
michael@0:   if (unlikely (script == USCRIPT_INVALID_CODE))
michael@0:     return HB_SCRIPT_INVALID;
michael@0: 
michael@0:   return hb_script_from_string (uscript_getShortName (script), -1);
michael@0: }
michael@0: 
michael@0: UScriptCode
michael@0: hb_icu_script_from_script (hb_script_t script)
michael@0: {
michael@0:   if (unlikely (script == HB_SCRIPT_INVALID))
michael@0:     return USCRIPT_INVALID_CODE;
michael@0: 
michael@0:   for (unsigned int i = 0; i < USCRIPT_CODE_LIMIT; i++)
michael@0:     if (unlikely (hb_icu_script_to_script ((UScriptCode) i) == script))
michael@0:       return (UScriptCode) i;
michael@0: 
michael@0:   return USCRIPT_UNKNOWN;
michael@0: }
michael@0: 
michael@0: 
michael@0: static hb_unicode_combining_class_t
michael@0: hb_icu_unicode_combining_class (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 				hb_codepoint_t      unicode,
michael@0: 				void               *user_data HB_UNUSED)
michael@0: 
michael@0: {
michael@0:   return (hb_unicode_combining_class_t) u_getCombiningClass (unicode);
michael@0: }
michael@0: 
michael@0: static unsigned int
michael@0: hb_icu_unicode_eastasian_width (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 				hb_codepoint_t      unicode,
michael@0: 				void               *user_data HB_UNUSED)
michael@0: {
michael@0:   switch (u_getIntPropertyValue(unicode, UCHAR_EAST_ASIAN_WIDTH))
michael@0:   {
michael@0:   case U_EA_WIDE:
michael@0:   case U_EA_FULLWIDTH:
michael@0:     return 2;
michael@0:   case U_EA_NEUTRAL:
michael@0:   case U_EA_AMBIGUOUS:
michael@0:   case U_EA_HALFWIDTH:
michael@0:   case U_EA_NARROW:
michael@0:     return 1;
michael@0:   }
michael@0:   return 1;
michael@0: }
michael@0: 
michael@0: static hb_unicode_general_category_t
michael@0: hb_icu_unicode_general_category (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 				 hb_codepoint_t      unicode,
michael@0: 				 void               *user_data HB_UNUSED)
michael@0: {
michael@0:   switch (u_getIntPropertyValue(unicode, UCHAR_GENERAL_CATEGORY))
michael@0:   {
michael@0:   case U_UNASSIGNED:			return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
michael@0: 
michael@0:   case U_UPPERCASE_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER;
michael@0:   case U_LOWERCASE_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER;
michael@0:   case U_TITLECASE_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER;
michael@0:   case U_MODIFIER_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER;
michael@0:   case U_OTHER_LETTER:			return HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER;
michael@0: 
michael@0:   case U_NON_SPACING_MARK:		return HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK;
michael@0:   case U_ENCLOSING_MARK:		return HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
michael@0:   case U_COMBINING_SPACING_MARK:	return HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK;
michael@0: 
michael@0:   case U_DECIMAL_DIGIT_NUMBER:		return HB_UNICODE_GENERAL_CATEGORY_DECIMAL_NUMBER;
michael@0:   case U_LETTER_NUMBER:			return HB_UNICODE_GENERAL_CATEGORY_LETTER_NUMBER;
michael@0:   case U_OTHER_NUMBER:			return HB_UNICODE_GENERAL_CATEGORY_OTHER_NUMBER;
michael@0: 
michael@0:   case U_SPACE_SEPARATOR:		return HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
michael@0:   case U_LINE_SEPARATOR:		return HB_UNICODE_GENERAL_CATEGORY_LINE_SEPARATOR;
michael@0:   case U_PARAGRAPH_SEPARATOR:		return HB_UNICODE_GENERAL_CATEGORY_PARAGRAPH_SEPARATOR;
michael@0: 
michael@0:   case U_CONTROL_CHAR:			return HB_UNICODE_GENERAL_CATEGORY_CONTROL;
michael@0:   case U_FORMAT_CHAR:			return HB_UNICODE_GENERAL_CATEGORY_FORMAT;
michael@0:   case U_PRIVATE_USE_CHAR:		return HB_UNICODE_GENERAL_CATEGORY_PRIVATE_USE;
michael@0:   case U_SURROGATE:			return HB_UNICODE_GENERAL_CATEGORY_SURROGATE;
michael@0: 
michael@0: 
michael@0:   case U_DASH_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_DASH_PUNCTUATION;
michael@0:   case U_START_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_OPEN_PUNCTUATION;
michael@0:   case U_END_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_CLOSE_PUNCTUATION;
michael@0:   case U_CONNECTOR_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_CONNECT_PUNCTUATION;
michael@0:   case U_OTHER_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_OTHER_PUNCTUATION;
michael@0: 
michael@0:   case U_MATH_SYMBOL:			return HB_UNICODE_GENERAL_CATEGORY_MATH_SYMBOL;
michael@0:   case U_CURRENCY_SYMBOL:		return HB_UNICODE_GENERAL_CATEGORY_CURRENCY_SYMBOL;
michael@0:   case U_MODIFIER_SYMBOL:		return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_SYMBOL;
michael@0:   case U_OTHER_SYMBOL:			return HB_UNICODE_GENERAL_CATEGORY_OTHER_SYMBOL;
michael@0: 
michael@0:   case U_INITIAL_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_INITIAL_PUNCTUATION;
michael@0:   case U_FINAL_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_FINAL_PUNCTUATION;
michael@0:   }
michael@0: 
michael@0:   return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
michael@0: }
michael@0: 
michael@0: static hb_codepoint_t
michael@0: hb_icu_unicode_mirroring (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 			  hb_codepoint_t      unicode,
michael@0: 			  void               *user_data HB_UNUSED)
michael@0: {
michael@0:   return u_charMirror(unicode);
michael@0: }
michael@0: 
michael@0: static hb_script_t
michael@0: hb_icu_unicode_script (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 		       hb_codepoint_t      unicode,
michael@0: 		       void               *user_data HB_UNUSED)
michael@0: {
michael@0:   UErrorCode status = U_ZERO_ERROR;
michael@0:   UScriptCode scriptCode = uscript_getScript(unicode, &status);
michael@0: 
michael@0:   if (unlikely (U_FAILURE (status)))
michael@0:     return HB_SCRIPT_UNKNOWN;
michael@0: 
michael@0:   return hb_icu_script_to_script (scriptCode);
michael@0: }
michael@0: 
michael@0: #if U_ICU_VERSION_MAJOR_NUM >= 49
michael@0: static const UNormalizer2 *normalizer;
michael@0: #endif
michael@0: 
michael@0: static hb_bool_t
michael@0: hb_icu_unicode_compose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 			hb_codepoint_t      a,
michael@0: 			hb_codepoint_t      b,
michael@0: 			hb_codepoint_t     *ab,
michael@0: 			void               *user_data HB_UNUSED)
michael@0: {
michael@0: #if U_ICU_VERSION_MAJOR_NUM >= 49
michael@0:   {
michael@0:     UChar32 ret = unorm2_composePair (normalizer, a, b);
michael@0:     if (ret < 0) return false;
michael@0:     *ab = ret;
michael@0:     return true;
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   /* We don't ifdef-out the fallback code such that compiler always
michael@0:    * sees it and makes sure it's compilable. */
michael@0: 
michael@0:   UChar utf16[4], normalized[5];
michael@0:   unsigned int len;
michael@0:   hb_bool_t ret, err;
michael@0:   UErrorCode icu_err;
michael@0: 
michael@0:   len = 0;
michael@0:   err = false;
michael@0:   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), a, err);
michael@0:   if (err) return false;
michael@0:   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), b, err);
michael@0:   if (err) return false;
michael@0: 
michael@0:   icu_err = U_ZERO_ERROR;
michael@0:   len = unorm_normalize (utf16, len, UNORM_NFC, 0, normalized, ARRAY_LENGTH (normalized), &icu_err);
michael@0:   if (U_FAILURE (icu_err))
michael@0:     return false;
michael@0:   if (u_countChar32 (normalized, len) == 1) {
michael@0:     U16_GET_UNSAFE (normalized, 0, *ab);
michael@0:     ret = true;
michael@0:   } else {
michael@0:     ret = false;
michael@0:   }
michael@0: 
michael@0:   return ret;
michael@0: }
michael@0: 
michael@0: static hb_bool_t
michael@0: hb_icu_unicode_decompose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 			  hb_codepoint_t      ab,
michael@0: 			  hb_codepoint_t     *a,
michael@0: 			  hb_codepoint_t     *b,
michael@0: 			  void               *user_data HB_UNUSED)
michael@0: {
michael@0: #if U_ICU_VERSION_MAJOR_NUM >= 49
michael@0:   {
michael@0:     UChar decomposed[4];
michael@0:     int len;
michael@0:     UErrorCode icu_err = U_ZERO_ERROR;
michael@0:     len = unorm2_getRawDecomposition (normalizer, ab, decomposed,
michael@0: 				      ARRAY_LENGTH (decomposed), &icu_err);
michael@0:     if (U_FAILURE (icu_err) || len < 0) return false;
michael@0: 
michael@0:     len = u_countChar32 (decomposed, len);
michael@0:     if (len == 1) {
michael@0:       U16_GET_UNSAFE (decomposed, 0, *a);
michael@0:       *b = 0;
michael@0:       return *a != ab;
michael@0:     } else if (len == 2) {
michael@0:       len =0;
michael@0:       U16_NEXT_UNSAFE (decomposed, len, *a);
michael@0:       U16_NEXT_UNSAFE (decomposed, len, *b);
michael@0:     }
michael@0:     return true;
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   /* We don't ifdef-out the fallback code such that compiler always
michael@0:    * sees it and makes sure it's compilable. */
michael@0: 
michael@0:   UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
michael@0:   unsigned int len;
michael@0:   hb_bool_t ret, err;
michael@0:   UErrorCode icu_err;
michael@0: 
michael@0:   /* This function is a monster! Maybe it wasn't a good idea adding a
michael@0:    * pairwise decompose API... */
michael@0:   /* Watchout for the dragons.  Err, watchout for macros changing len. */
michael@0: 
michael@0:   len = 0;
michael@0:   err = false;
michael@0:   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), ab, err);
michael@0:   if (err) return false;
michael@0: 
michael@0:   icu_err = U_ZERO_ERROR;
michael@0:   len = unorm_normalize (utf16, len, UNORM_NFD, 0, normalized, ARRAY_LENGTH (normalized), &icu_err);
michael@0:   if (U_FAILURE (icu_err))
michael@0:     return false;
michael@0: 
michael@0:   len = u_countChar32 (normalized, len);
michael@0: 
michael@0:   if (len == 1) {
michael@0:     U16_GET_UNSAFE (normalized, 0, *a);
michael@0:     *b = 0;
michael@0:     ret = *a != ab;
michael@0:   } else if (len == 2) {
michael@0:     len =0;
michael@0:     U16_NEXT_UNSAFE (normalized, len, *a);
michael@0:     U16_NEXT_UNSAFE (normalized, len, *b);
michael@0: 
michael@0:     /* Here's the ugly part: if ab decomposes to a single character and
michael@0:      * that character decomposes again, we have to detect that and undo
michael@0:      * the second part :-(. */
michael@0:     UChar recomposed[20];
michael@0:     icu_err = U_ZERO_ERROR;
michael@0:     unorm_normalize (normalized, len, UNORM_NFC, 0, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
michael@0:     if (U_FAILURE (icu_err))
michael@0:       return false;
michael@0:     hb_codepoint_t c;
michael@0:     U16_GET_UNSAFE (recomposed, 0, c);
michael@0:     if (c != *a && c != ab) {
michael@0:       *a = c;
michael@0:       *b = 0;
michael@0:     }
michael@0:     ret = true;
michael@0:   } else {
michael@0:     /* If decomposed to more than two characters, take the last one,
michael@0:      * and recompose the rest to get the first component. */
michael@0:     U16_PREV_UNSAFE (normalized, len, *b); /* Changes len in-place. */
michael@0:     UChar recomposed[18 * 2];
michael@0:     icu_err = U_ZERO_ERROR;
michael@0:     len = unorm_normalize (normalized, len, UNORM_NFC, 0, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
michael@0:     if (U_FAILURE (icu_err))
michael@0:       return false;
michael@0:     /* We expect that recomposed has exactly one character now. */
michael@0:     if (unlikely (u_countChar32 (recomposed, len) != 1))
michael@0:       return false;
michael@0:     U16_GET_UNSAFE (recomposed, 0, *a);
michael@0:     ret = true;
michael@0:   }
michael@0: 
michael@0:   return ret;
michael@0: }
michael@0: 
michael@0: static unsigned int
michael@0: hb_icu_unicode_decompose_compatibility (hb_unicode_funcs_t *ufuncs HB_UNUSED,
michael@0: 					hb_codepoint_t      u,
michael@0: 					hb_codepoint_t     *decomposed,
michael@0: 					void               *user_data HB_UNUSED)
michael@0: {
michael@0:   UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
michael@0:   unsigned int len;
michael@0:   int32_t utf32_len;
michael@0:   hb_bool_t err;
michael@0:   UErrorCode icu_err;
michael@0: 
michael@0:   /* Copy @u into a UTF-16 array to be passed to ICU. */
michael@0:   len = 0;
michael@0:   err = FALSE;
michael@0:   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), u, err);
michael@0:   if (err)
michael@0:     return 0;
michael@0: 
michael@0:   /* Normalise the codepoint using NFKD mode. */
michael@0:   icu_err = U_ZERO_ERROR;
michael@0:   len = unorm_normalize (utf16, len, UNORM_NFKD, 0, normalized, ARRAY_LENGTH (normalized), &icu_err);
michael@0:   if (icu_err)
michael@0:     return 0;
michael@0: 
michael@0:   /* Convert the decomposed form from UTF-16 to UTF-32. */
michael@0:   icu_err = U_ZERO_ERROR;
michael@0:   u_strToUTF32 ((UChar32*) decomposed, HB_UNICODE_MAX_DECOMPOSITION_LEN, &utf32_len, normalized, len, &icu_err);
michael@0:   if (icu_err)
michael@0:     return 0;
michael@0: 
michael@0:   return utf32_len;
michael@0: }
michael@0: 
michael@0: 
michael@0: hb_unicode_funcs_t *
michael@0: hb_icu_get_unicode_funcs (void)
michael@0: {
michael@0:   static const hb_unicode_funcs_t _hb_icu_unicode_funcs = {
michael@0:     HB_OBJECT_HEADER_STATIC,
michael@0: 
michael@0:     NULL, /* parent */
michael@0:     true, /* immutable */
michael@0:     {
michael@0: #define HB_UNICODE_FUNC_IMPLEMENT(name) hb_icu_unicode_##name,
michael@0:       HB_UNICODE_FUNCS_IMPLEMENT_CALLBACKS
michael@0: #undef HB_UNICODE_FUNC_IMPLEMENT
michael@0:     }
michael@0:   };
michael@0: 
michael@0: #if U_ICU_VERSION_MAJOR_NUM >= 49
michael@0:   if (!hb_atomic_ptr_get (&normalizer)) {
michael@0:     UErrorCode icu_err = U_ZERO_ERROR;
michael@0:     /* We ignore failure in getNFCInstace(). */
michael@0:     hb_atomic_ptr_cmpexch (&normalizer, NULL, unorm2_getNFCInstance (&icu_err));
michael@0:   }
michael@0: #endif
michael@0:   return const_cast<hb_unicode_funcs_t *> (&_hb_icu_unicode_funcs);
michael@0: }
michael@0: 
michael@0: